B.AI Upgrades Model Discount Matrix: GLM-5.3-Flash Now 40% Off, Hy4 at 50% Off

Key Info

B.AI updated its model discount matrix, lowering GLM-5.3-Flash from 30% to 40% off while offering Tencent Hunyuan Hy4 at 50% off (limited time), DeepSeek-V4.1-Flash-1017 at 40% off, and Qwen3.8-Flash at 20% off through a single unified API with intelligent routing.

Highlights

• GLM-5.3-Flash: 320B native multimodal model, now 40% off, aimed at coding, visual programming, and complex tool calling for agent workflows.
• Tencent Hunyuan Hy4: 50% off until Dec 31, 2026 (SGT), adds visual understanding, text/image input, and a 1M-token context window.
• DeepSeek-V4.1-Flash-1017: 763B open model at 40% off, 1M context, interleaved text-image input, optimized for DeepSeek Harness.
• Qwen3.8-Flash: 20% off, hosted 1M context window with text, image, and video multimodal input.
• B.AI positions itself as a settlement/routing layer above models and beneath agents, letting developers switch flagship models with one integration.